Research On Service Based Radio Access Network Architecture And Radio Resource Allocation Algorithm

With the commercialization of the Fifth Generation Mobile Communication System(5G)technology and the development of the Sixth Generation Mobile Communication System(6G)technology,new services and application scenarios such as the Internet of Things and vertical industries are increasingly rich,resulting in increasingly diversified business requirements and stringent performance indicators.In order to solve the above problems,the Service Based Architecture(SBA)is proposed for 5G core network,which further divides the network elements of the traditional core point-to-point communication into network functions,so as to realize the rapid deployment,flexible configuration,and management of the core network.In addition,network functions can be flexibly combined according to service characteristics to realize core network slice and meet the differentiated needs of different services.However,due to the complex wireless environment and diversified application scenarios of the Radio Access Network(RAN),the open design of the its architecture is more challenging.In addition,dynamic changes in business requirements make it difficult to customize RAN slice based on open architecture research.Therefore,it is urgent to study service-based RAN architecture and RAN slice resource dynamic allocation algorithm.To solve these problems,this paper studies service-based RAN architecture and wireless resource allocation algorithm.First of all,a "four layers and three sides" service-based RAN architecture was designed.This architecture has the characteristics of flexible deployment and easy function expansion,which can support the efficient allocation of wireless network resources on demand.Secondly,in view of the differentiation of key performance indicators of slicing caused by the diversification of network slicing and the tradeoff between slicing performance and operating cost,we explored the joint optimization of multi-dimensional performance based on deep reinforcement learning and the allocation of time-frequency resources of computational cost,which could support the flexible configuration and management of service-based RAN.The research content of this paper is as follows:(1)This paper proposes a service-based RAN architecture and implements the platform.Firstly,the RAN is abstracted as "four layers and three planes",in which the "four layers" from top to bottom are the application layer,the function layer,the virtualization layer,and the infrastructure layer,while the "three planes" include the user plane,the control plane and the management and orchestration plane.Wireless network infrastructure can be flexibly reconfigured and managed by virtualization and introducing programmability to wireless networks.Secondly,the architecture draws lessons from SBA and micro-service ideas to carry out service-based design.Based on the characteristics of RAN,its control plane is decoupled into radio resource control function,policy configuration function,Quality of Service(Qo S)analysis function and status monitoring function.The Hypertext Transfer Protocol(HTTP)is used to obtain and provide services among functions.Then,according to the requirements of different types of network applications for radio resources,transmission parameters and transmission mechanisms,the corresponding templates are designed.Kubernetes is used to implement template instantiation,so as to realize the ondemand reconstruction and management of service-based RAN.Finally,a service-based RAN experimental verification platform is built on a general server based on open source software.The test analysis shows that the proposed architecture can improve service update speed and response speed,and reduce the signaling overhead and the execution time of resource configuration command.(2)As for both the enhanced Mobile Broadband(e MBB)slice for high-speed transmission requirements and Ultra Reliable and Low Latency Communication(URLLC)slice for lowlatency requirements,this paper jointly optimize the multidimensional performance indicator of slice and the operating cost of base station,and construct the system cumulative utility function maximization problem.Under the constraints that the wireless resources allocated to users should not exceed the total bandwidth resources,the URLLC slice rate should not be less than the lower limit of rate,and the e MBB slice service delay should not be more than the upper limit of the service delay,the optimized time-frequency resource allocation is adopted to support the flexible resource allocation of service-based RAN.In this paper,the performance differences and operating costs of different slice types are considered to allocate wireless resources,and the utility function is defined as the weighted sum of the rate benefit of e MBB slice,the delay benefit of URLLC slice,and the computing resource cost of the base station.Secondly,because the traditional optimization algorithm is difficult in solving the long-term optimization objective with complex and difficult calculation.The optimization objective is modeled as Markov decision process and solved by Double Deep Q Network(DDQN)algorithm.Combined with constraints,the system utility function is taken as the conditional reward.Maximize cumulative rewards by dynamically allocating limited time-frequency resources over a physical network.Finally,the simulation results show that compared with the utility optimization resource allocation algorithm based on DQN,the traditional single performance optimization resource allocation algorithm and static uniform allocation algorithm,the utility optimization resource allocation algorithm based on DDQN designed in this paper can achieve higher bandwidth utilization and lower computing energy consumption while satisfying the Qo S of users.